Language selection

Search

Processing Artificial Intelligence: Dataset

From: Canadian Intellectual Property Office


Dataset

The breakdown of the dataset used for this report, which covers the 1998-2017 period, is presented in Figure 1. The worldwide dataset is comprised of approximately 85,000 AI patented inventions. Combined Canadian researchers and institutions account for 1.8% of the international dataset, or more than 1,500 patented inventions. Our analysis divides this dataset into two subsets by separately examining the data from the perspective of filings by Canadian institutions and another focused on filings by researchers. The institutions dataset is comprised of 284 Canadian institutions, which, for the purpose of this analysis, includes academic institutions, companies and government departments. In total, these Canadian institutions patented more than 600 inventions. The researcher dataset consists of approximately 2,500 Canadian inventors that are associated with more than 1,400 patented inventions.

Figure 1: Breakdown of AI patented inventions by dataset analyzed in this report

Description of figure 1

Figure 1 is a diagram in the shape of an upside down triangle that forks at the tip with four partitions showing a breakdown of patented inventions by data set. The upper partition shows that worldwide patented inventions in AI were 85,144. This partition is shaded in dark blue with text in white. The middle partition shows that worldwide patented inventions in AI attributable to a Canadian firm or research institution were 1,516. This partition is shaded in light blue with text in white. Beside this partition, it is written that there are 284 Canadian institutions and 2,497 Canadian researchers in this data set. The lower left-hand side partition of the reversed triangle tip shows that the Canadian Institution data set consists of 618 patented inventions and the lower right-hand partition of the reversed triangle tip shows that the Canadian Researcher data set consists of 1,419 patented inventions. These two partitions are shaded in light green with values in white. Overlaid on these two partitions is text written in white that states “Worldwide AI Patented Inventions”. Beside the two partitions, two labels indicate the Canadian Institution data set and Canadian Researcher data set.

  • Worldwide Patented Inventions in Artificial Intelligence: 85,144
  • Worldwide AI Patented Inventions Attributable to a Canadian Institution or Researcher: 1,516

Worldwide AI Patented Inventions Data

  • Number of Canadian Institutions: 284
  • Number of Canadian Researchers: 2,497
  • Number of Patented Inventions in Canadian Institution data set: 618
  • Number of Patented Inventions in Canadian Researcher data set: 1,419

In order to gain a deeper understanding of where Canada’s strengths lie in AI and how it fares in comparison with other nations, the international dataset is further broken down using the taxonomy presented by WIPO in the Technology Trends 2019 report.Footnote 10 The taxonomy, which is based on the Association for Computing Machinery (ACM) Computing Classification System, involves grouping the data into three categories: AI Techniques, AI Applications and AI Fields. The AI Techniques category includes advanced forms of statistical models that facilitate the computation of tasks typically performed by humans; the AI Applications category includes functions that can be realized using one or more AI Techniques; and the AI Fields category includes the different fields, areas or disciplines where AI Techniques or AI Applications may be applied. Breaking down the AI patented inventions dataset into these three categories is effective from an analytical perspective, as each category represents developments in AI from different perspectives, moving from the fundamental technical aspects of AI, to applications of AI at the functional level or as used in specific industries. The distribution of this data across these three AI categories is presented in Figure 2. These AI categories are established based on the inclusion of a combination of International Patent Classification (IPC) codes, Cooperative Patent Classification (CPC) codes and a set of relevant keywords in each patented invention.Footnote 11 In assigning the patented inventions to each of these designated categories, we note that not all patented inventions are classified according to these specific groupings. An area of potential future work would be to analyze these patented inventions to determine whether they are representative of emerging fields of AI that have not yet been classified.

The Canadian dataset is categorized using the same approach as the international dataset. One of the most noticeable differences between the two datasets is that a larger proportion of Canadian inventions is categorized under AI Fields only, as opposed to both AI Fields and AI Techniques. On the other hand, a larger percentage of patented inventions in Canada are categorized to both AI Techniques and AI Applications rather than just AI Applications. The intersection of AI Applications and AI Fields is lower in the Canadian dataset compared to the international dataset. There are no significant differences for the intersection of all three categories, as shown by the inner circle of each diagram, or for the outer circle labelled as the “Other” category.

Figure 2: Distribution of patented inventions across AI categories for International and Canadian AI datasets

Description of figure 2

Figure 2 consists of two graphs placed side by side. The graph on the left describes the International AI data set and the one on the right describes the Canadian AI data set. Each graph consists of three circles stacked on top of each other. The smallest circle in the centre shows the intersection of all three categories labelled on the graph (A – AI Techniques, B – AI Applications, and C – AI Fields. The second circle behind the circle in the middle is divided into six regions. Three of these sections show the number of inventions for each of the three categories. The three sections in between these three categories represent the data for the intersection between the two categories beside it. The largest circle behind the two smaller circles represents the "Other" category. This category represents patented inventions that do not belong to any of the three categories.

International AI data set 
Grouping Patented inventions Percentage

A - AI Techniques

14,922

17.5%

B - AI Applications

4,069

4.8%

C - AI Fields

12,804

15.0%

A & B

6,704

7.9%

A & C

18,921

22.2%

B & C

4,927

5.8%

A, B & C

12,272

14.4%

Other

10,525

12.4%

Canadian AI dataset
Grouping Patented inventions Percentage

A - AI Techniques

290

19.1%

B - AI Applications

59

3.9%

C - AI Fields

290

19.1%

A & B

156

10.3%

A & C

297

19.6%

B & C

70

4.6%

A, B & C

185

12.2%

Other

169

11.1%

Date modified: